Six Sigma Improvement Solidification in Practice: Two Case Studies from Effect Verification to Standardized Control

1. Introduction: Why Do 70% of Improvement Projects "Revert" Within Six Months?

In early 2025, a consulting firm conducted a follow-up survey on Six Sigma projects in 87 domestic manufacturing companies, revealing a shocking statistic: over 70% of projects that had completed the entire DMAIC process and passed phase gate reviews experienced a decline in effectiveness within six months after acceptance—yield rates regressed, defect rates increased, and pre-improvement issues reappeared in new forms. This phenomenon is known in the industry as "improvement reversion" or "result drift."

The root cause does not lie in the ineffectiveness of the improvement solutions themselves, but rather in the severe underestimation and oversimplification of the final stage of DMAIC—Control (控制). Many Six Sigma project teams, after finding solutions and verifying their effects in the Improve stage, hastily declare the project complete, reducing the control stage to "updating a control plan" or "writing a work instruction." This hasty conclusion robs the improvement results of a solid foundation for long-term maintenance.

In fact, Bill Smith, one of the founders of Six Sigma and Motorola's chief statistician, repeatedly emphasized a key point: "The true value of improvement lies not in the moment it occurs, but in the day when the improvement effects can be sustained." The core mission of the Control stage is to embed the improvement results into the organization's daily operational system through a systematic mechanism—standardization, poka-yoke, process monitoring, training solidification, and emergency response—so that the results do not depend on the personal will of project team members but become a system capability that operates automatically regardless of who is involved.

This article will present two real-world case studies to fully illustrate the practical path of solidifying improvement results.

2. Case Study One: A Welding Process in an Automotive Parts Company Improves from 87% to 99.2% Yield Rate

2.1 Project Background and Improvement Results

In March 2024, a joint venture automotive parts company—Jianglin Precision (化名)—faced a significant quality challenge on one of its chassis welding lines. This line produces the rear subframe assembly for a mainstream SUV model, involving two key processes: robotic arc welding and manual spot welding. Before the project started, the first-time welding pass rate was only 87.3%, with the main defects being porosity (42%), spatter (28%), and weld bead (19%). The monthly rework hours exceeded 320, and the customer PPM was as high as 2,850.

Through the first four stages of DMAIC, the Black Belt team identified three critical root causes: unstable protective gas flow (due to aging gas supply lines for the welding robots), weld wire extension length deviation exceeding the specification range, and excessive surface oil contamination in incoming materials. The team implemented improvement measures for these root causes—replacing the gas supply lines and installing real-time gas flow monitoring devices, creating a weld wire extension length positioning fixture, and adding a degreasing cleaning process before the welding operation. After verification, the first-time welding pass rate increased to 99.2%, with a 91% reduction in porosity defects, and the PPM dropped to 180.

However, the real test was just beginning.

2.2 Standardization: From "Person-to-Person Monitoring" to "System Management"

When the project entered the Control stage, the quality director posed a sharp question: "If the operator changes in three months, or the equipment maintenance personnel forget to replace the gas supply lines, will the pass rate fall back?"

This question hit the core. The Black Belt team developed a three-tier standardization strategy.

The first tier is the standardization of process parameters. The team locked all critical process parameters of the welding robots—protective gas flow range (15～18L/min), welding speed (450～550mm/min), wire feeding speed (6.5～7.2m/min), and weld wire extension length (12～15mm)—into the robot control program. The parameter modification permissions were set to "adjustable only by process engineers," and any parameter change must go through a change management process. This ensures that operators cannot arbitrarily adjust parameters, fundamentally preventing human errors.

The second tier is the standardization of work actions. For the manual spot welding process, the team conducted a motion study and developed a standardized work combination table (Standardized Work Combination Table), which detailed the sequence, time, and precautions of each spot welding action in a visual format on the kanban at the workstation. Additionally, the team produced standard work videos to serve as training materials for new operators and annual retraining.

The third tier is the standardization of maintenance. The team collaborated with the equipment maintenance department to create a regular inspection checklist for the welding gas supply lines, requiring operators to check and record the gas flow meter readings before each shift. Equipment maintenance personnel must perform a gas tightness test weekly and replace the filter core of the gas supply lines monthly. All inspection records are integrated into the equipment management system, and the system automatically sends reminders to the shift leader and equipment supervisor if the records are overdue.

2.3 Poka-Yoke: Making Abnormalities "Impossible to Occur" is More Reliable than "Detecting Them"

Standardization addresses the "how to do it" issue, but human negligence is always difficult to completely avoid. Therefore, the team introduced poka-yoke devices at critical control points.

One of the most successful examples is the real-time gas flow monitoring and alarm system. The team installed an electronic gas flow sensor at the gas inlet of the welding robot, connecting the flow signal to the robot controller. When the flow rate drops below 15L/min or exceeds 18L/min, the robot automatically stops the welding action and triggers an audio-visual alarm, displaying a fault code on the central control screen. This poka-yoke device fundamentally prevents the issue of continuing to weld when the gas flow is unstable, reducing the risk of porosity defect recurrence to nearly zero.

Another poka-yoke measure was for the weld wire extension length. The original method relied on the operator's experience to visually judge the length, often leading to deviations. The team created an L-shaped length positioning fixture—when the welding gun nozzle is inserted into the positioning slot of the fixture, the tip of the weld wire just touches the fixture's limit surface, ensuring the extension length is the standard value of 13.5mm. Operators only need to perform the "insert—confirm" action once each time they change the wire spool, a process that takes no more than 3 seconds and completely eliminates judgment errors.

2.4 Process Monitoring: Let Data Tell You to Take Action Before It's Too Late

One of the core tools in the Control stage is statistical process control (SPC). The team established control charts for three key quality characteristics (CTQ) of the welding process: welding current control chart (Xbar-R), protective gas flow control chart (I-MR), and post-weld porosity count control chart (C chart).

To prevent the control charts from becoming mere "monthly Excel sheets," the team integrated them into the shop floor MES system—real-time process parameters of the welding robot are automatically collected every 5 seconds, and the MES system automatically plots the data and runs the eight criteria for identifying out-of-control conditions. Once an abnormal signal is triggered, the system immediately sends a warning to the process engineer's mobile terminal and locks the product batch at that workstation, requiring confirmation by the process personnel before release.

This digital monitoring approach proved highly effective. During the 12-month follow-up period after the improvement, the system triggered 37 warnings, 31 of which were false alarms (Type I Error), and the remaining 6 were genuine abnormalities. Among these 6 genuine abnormalities, 4 were detected and handled within 30 minutes of occurrence, and only 2 resulted in a small number of nonconforming products—3 and 5 pieces, respectively, far below the pre-improvement average of about 35 pieces per shift.

2.5 Data on Sustaining Results

As of the 18th month after project acceptance, the first-time welding pass rate on Jianglin Precision's chassis welding line stabilized between 98.7% and 99.5%, with a monthly fluctuation of no more than 0.8 percentage points. The annual financial benefits generated by the project were approximately 2.47 million yuan (reduced rework costs + reduced scrap + reduced customer penalties), exceeding the 2.1 million yuan estimated in the project Charter. This case has since become a benchmark for "improvement result solidification" within the company.

3. Case Study Two: SMT Line Component Misalignment Improvement and Horizontal Expansion in an Electronics Company

3.1 Project Background

In July 2024, a consumer electronics contract manufacturer—Huari Electronics (化名)—faced a severe component misalignment issue on one of its high-speed SMT lines. This line primarily assembles the motherboard for a brand of smartphones, involving over 400 components, with the smallest being 0201 (0.6mm×0.3mm) passive components. At the start of the project, the component misalignment defect rate was as high as 1,850ppm, 3.7 times the customer's allowed upper limit of 500ppm. More challenging was the intermittent nature of the issue—sometimes there were zero defects for several hours, and other times a large number of misalignments suddenly appeared, causing a sharp increase in AOI false alarm rates and frequent line stops for verification.

After thorough investigations in the Measure and Analyze stages, the Green Belt team identified the key influencing factors for component misalignment: reduced stencil tension (stencil usage exceeding 30,000 times results in tension below 35N/cm), abnormal collapse of solder paste shape after printing, and excessive wear of the pick-and-place machine nozzles beyond their service life. The team validated the optimal parameter combination through DOE and implemented the improvement measures: establishing a stencil tension management ledger and proactively replacing stencils before reaching their service life, optimizing solder paste printing parameters and adding 3D SPI (solder paste inspection) process control, and establishing a nozzle life management system.

After implementing the improvement measures, the component misalignment defect rate dropped from 1,850ppm to 220ppm, and the line passed the customer's quality validation. However, the Green Belt team faced the true challenge: how to replicate and sustain this result across three different SMT lines?

3.2 Control Plan and Process Control

The first task of the team in the Control stage was to develop a complete control plan (Control Plan), systematically organizing all process parameters, control methods, and response plans that needed to be controlled after the improvement.

Key points of the control plan include:

• Stencil Tension: Control standard ≥ 40N/cm, inspection frequency once every 5,000 uses, control method using a tension tester, and a response process of "stop use—replace stencil—trace printed PCBs" if the specification is exceeded.
• Solder Paste Printing Thickness: Control standard 90%~110% of the stencil thickness (stencil thickness 0.12mm, target thickness 0.108～0.132mm), inspection frequency every 2 hours or every 50 PCBs, control method using 3D SPI online inspection, and a response plan of "immediate shutdown—adjust printing parameters—trace the previous 50 PCBs" if the specification is exceeded.
• Nozzle Condition: Vacuum and placement accuracy calibration using a nozzle checker before each shift, and immediate replacement if the deviation exceeds ±10%.

This control plan is not just a tool for the quality department; it is included as a mandatory review item in the daily production meetings—production supervisors must report the compliance rate of all CTQs at the morning meeting, and any parameter exceeding the control limits is marked as a "red anomaly" and must be analyzed and corrected by the process engineer within 24 hours.

3.3 Training Matrix and Personnel Certification

The biggest resistance in the control stage came from the skill differences among operators. New and experienced operators have different habits and varying understandings of control requirements. To address this, the team established a welding quality control training matrix (Training Matrix), breaking down the operation requirements for each process into four dimensions: theoretical training, practical training, assessment certification, and recertification cycle.

Specific actions include:

1. Writing standardized training materials for each workstation on the SMT line (solder paste printing, pick-and-place machine operation, reflow welding monitoring, AOI inspection), covering three modules: process parameter ranges, defect identification, and emergency handling procedures.
2. Establishing a three-tier skill certification system: Green Card for "independent operation," requiring a theoretical exam score of 80% or higher and a practical assessment of processing 100 PCBs without quality issues; Blue Card for "training new employees," requiring the ability to identify common defects and handle them independently; Gold Card for "cross-workstation operation," requiring certification in all related workstations.
3. Conducting recertification every six months, with those who fail being retrained before resuming their positions.

After implementing this system, the skill differences among SMT line operators were significantly reduced, and the process capability index (Cpk) of the line steadily improved from 0.87 to 1.56, with no quality fluctuations due to operator differences.

3.4 Horizontal Expansion: From One Line to Factory-Wide Replication

Huari Electronics has a total of 5 SMT lines, with Line 1 completing the improvement, but the other 4 lines still using the old process. If the component misalignment issue on these 4 lines is not resolved, the company's overall quality performance will still fail to meet customer standards.

The team adopted a "standard package replication" method for horizontal expansion. Specifically, they bundled the standardized outputs from Line 1—control plan, poka-yoke device diagrams, training materials, SPC control limits, and MES monitoring configurations—into an "improvement result replication package" and directly applied it to Lines 2 to 5. For each line, a team of three people, consisting of a process engineer and a trainer, was dispatched to complete deployment, debugging, and operator training on-site for two weeks.

This replication strategy was highly efficient. The deployment for Line 2 took 21 days, Line 3 was shortened to 15 days, and by Line 5, the entire process was completed in just 10 days. By March 2025, the component misalignment defect rates on all 5 SMT lines were controlled below 300ppm, providing a solid foundation for achieving the company's annual quality target (component misalignment ≤ 500ppm across all lines).

3.5 Data on Sustaining Results

As of the 12th month after project acceptance, the monthly average component misalignment defect rate on Huari Electronics' 5 SMT lines stabilized between 180ppm and 260ppm. The annual quality benefits generated by the project were approximately 3.86 million yuan (reduced scrap + reduced customer rework penalties + reduced inspection costs), far exceeding the 2.8 million yuan estimated in the project Charter. The project was rated as the best Six Sigma project of the year, and the "improvement result replication package" methodology was promoted to other product line improvement projects.

4. Five Core Mechanisms for Solidifying Improvement Results

From the above two case studies, we can distill five core mechanisms for solidifying improvement results, which form the operational framework for the Control stage of DMAIC.

Mechanism One: Standardization—From "Doing It Right" to "Always Doing It Right." The essence of standardization is not just writing work instructions but locking the best practices after improvement into the organization's "default mode." Whether it is locking process parameters, standardizing work actions, or fixing maintenance cycles, the ultimate goal of standardization is to make the correct way the easiest way.

Mechanism Two: Poka-Yoke—Making Errors "Impossible to Occur" is More Reliable than "Detecting Them Timely." Poka-yoke is the most powerful tool in the Control stage. A well-designed poka-yoke device can automatically intercept errors when the operator is negligent, fatigued, or inexperienced. In resource-limited situations, prioritizing investment in poka-yoke devices often yields a higher return on investment than increasing inspection frequency.

Mechanism Three: Monitoring—Using Data to Build the "First Line of Defense" for Abnormality Warnings. Control charts, SPC, online inspection equipment, and MES warnings—these monitoring tools are not for post-event statistics but for real-time alerts. The monitoring system in the Control stage should have three characteristics: real-time (detecting abnormalities as soon as they occur), traceability (tracing to specific batches and times), and closed-loop (having associated response and confirmation processes after triggering an alert).

Mechanism Four: Humanization—Embedding Improvement Results into Organizational Capabilities Through Training and Certification. Even the best control plans and poka-yoke devices can be "cleverly bypassed" if operators do not understand their principles or recognize their value. Humanization tools such as training matrices, certification systems, and mentorship ensure that improvement results are not just written policies but ingrained behavior habits in employees' minds.

Mechanism Five: Institutionalization—Incorporating Improvement Results into Daily Management Systems. The execution status of the control plan should be a mandatory review item in daily production meetings, process capability indicators should be included in departmental KPIs, and control plan updates should be part of the change management process. Only when improvement results are integrated into the organization's daily management rhythm can they truly achieve "permanent solidification."

These five mechanisms support each other and are indispensable. Standardization provides the foundation, poka-yoke provides the guarantee, monitoring provides the early warning, humanization provides the capability, and institutionalization provides the sustainability. A complete Control stage plan should include all five dimensions.

5. Conclusion: The End of Improvement is Not "Achieving It," but "Making It Unstoppable"

In coaching Six Sigma projects, I often hear project teams say, "The improvement solution has been verified, and the project is complete." This mindset is dangerous. The true dividing line in a Six Sigma project is not whether a solution was found in the Improve stage, but whether a system has been established in the Control stage to sustain the improvement effects even after the project team is disbanded.

Referring back to the survey data mentioned at the beginning of this article—those 70% of projects that experienced a decline in effectiveness within six months—almost all of them made the same mistake: equating the Control stage with "updating documents." The two projects that successfully maintained their results for 18 months and 12 months—Jianglin Precision and Huari Electronics—were teams that invested as much or even more effort in the Control stage as they did in the improvement stage.

The solidification of improvement results is essentially a transition from "relying on individual capabilities" to "organizational system capabilities." When the output of an improvement project no longer depends on a specific person, position, or team's continuous attention but is embedded in the organization's processes, systems, culture, and incentive mechanisms, the improvement can be considered truly complete.

As the core philosophy of the Toyota Production System reveals: continuous improvement is not a project but a production method. Making improvement "unstoppable" is the highest realm of the Six Sigma Control stage.

The true value of improvement lies not in the moment it occurs, but in the day when the improvement effects can be sustained.

Knowledge code: 6.1.1

Version: v20260730

Author: Quality Think Tank Quality Think Tank is dedicated to providing systematic professional knowledge, methodologies, and practical tools for quality management practitioners, helping companies continuously enhance their quality capabilities.